Conventionally, for system monitoring, thresholds are set for a server to be monitored and monitoring items capable of being monitored by the server and the thresholds are monitored for the monitoring items, respectively, thereby detecting an abnormality. However, a proper threshold for each monitoring item is difficult to set and a setting work load is heavy, and thus PTL 1 discloses a technique “comprising a load model data creating means 14 for creating load model data indicating a temporal transition of system loads based on past load information of a computer system 1, a threshold data creating means 16 for adding designated threshold correction data to the load model data and time-sequentially calculating threshold data, and an abnormal load detecting means 17 for detecting an abnormal load of the system by comparing current load information of the system 1 with threshold data at a corresponding time thereto.